(165 days)
The Micro C Medical Imaging System, M01 is a handheld and portable general purpose X-ray system that is indicated for use by qualified/trained clinicians on adult and pediatric patients for taking diagnostic static and serial radiographic exposures of extremities. The device is not intended to replace a radiographic system that has both variable tube current and voltages (kVp) in the range that may be required for full optimization of image quality and radiation exposure for different exam types.
The Micro C Medical Imaging System, M01 (subject device) is a handheld X-ray system designed to aid clinicians with point of care visualization through diagnostic X-rays of distal extremities. The device allows a clinician to select desired technique factors best suited for their patient anatomy. The Micro C Medical Imaging System, M01 consists of three major subsystems: The Emitter, Cassette, and Control Unit. The System is intended to interface an external Monitor (touchscreen or non-touchscreen display), keyboard and a mouse, and can provide a remote operator interface over the network to a laptop. The Micro C Medical Imaging System, M01 utilizes a computer vision positioning system to allow the emitter to be positioned above the patient anatomy and aligned to the cassette by the operator. The device is used in a clinical environment.
The provided document describes the Micro C Medical Imaging System, M01, and updates to its software to include AiLARA (Artificial Intelligence-based Algorithm for Radiography) modes. The primary focus of the document regarding acceptance criteria and performance relates to the validation of this new software feature.
Here's a breakdown of the requested information:
1. A table of acceptance criteria and the reported device performance
The document does not explicitly state quantitative acceptance criteria for the AiLARA algorithm's performance in terms of image quality or diagnostic accuracy using specific metrics like sensitivity, specificity, or AUC. Instead, the acceptance criteria seem to be qualitative and focused on the algorithm's learning trend, software requirement fulfillment, diagnostic relevance of images, and radiation dose limits.
| Acceptance Criteria Category | Description of Acceptance Criteria (Inferred) | Reported Device Performance |
|---|---|---|
| AiLARA Algorithm Verification | The model should learn the trend of the training dataset (truth). Mean Squared Error (MSE) and Mean Absolute Error (MAE) for both training and testing datasets should indicate no further benefit from additional training (epochs). | The model was able to learn the trend of the training dataset (truth). The mean squared error of the training and verification testing datasets were plotted, and the trend lines showed that the model had learned the general trend present in the data. Both training and testing Mean Squared Error and Mean Absolute Error showed that additional training (epochs) would have no added benefit. |
| Software Verification | The updated Micro C software should meet system-level software requirements, and software outputs should meet expected results. | Software outputs met the expected result in all cases, with no anomalies found. |
| Image Quality Validation | Images generated with AiLARA modes should be diagnostically and clinically relevant when reviewed by board-certified radiologists and an orthopedic surgeon. | All images were determined to be diagnostically and clinically relevant. |
| Radiation Dose Testing | AiLARA mode's radiation outputs should be below established Diagnostic Reference Levels (DRLs) and not statistically different from the predicate device's manual mode dose outputs for comparable techniques. | All AiLARA dose values were below the established Diagnostic Reference Levels (DRLs) and there was no statistical difference between AiLARA and Manual mode calculated entrance skin exposure doses. |
| Radiation Dose Testing on Small/Pediatric Anatomies | AiLARA's radiation outputs for small size extremity anatomies should be below DRLs and consistent across various emitter orientations and small anatomy thicknesses. | All AiLARA dose values were below the established Diagnostic Reference Levels (DRLs) for small size anatomies, and doses were similar among captures for each orientation within the same target thickness and SID category. |
| Usability Evaluation | All critical use tasks for the AiLARA modes should be completed with a passing result by 100% of participants. | All the identified critical use tasks were completed with a passing result by 100% of participants. The usability evaluation was performed in accordance with IEC 62366-1:2020 and FDA guidance. |
2. Sample size used for the test set and the data provenance
- Test Set Sample Size: The AiLARA algorithm's full development dataset was split, with 20% of the data used as the testing set for algorithm verification.
- Data Provenance: The document does not specify the country of origin. The data used for algorithm verification was part of the "full development dataset" of AiLARA. For the "Image Quality Validation Study," the validation set was collected after the algorithm was frozen and transferred to the device, using phantoms at different emitter orientations and angles. The phantom types included ankle, elbow, hand, foot, knee, toe, and wrist. The study appears to be prospective in the sense that images were collected specifically for validation after the algorithm's finalization.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts
- For the Image Quality Validation Study, images were reviewed and rated by board certified radiologists and an orthopedic surgeon. The exact number of experts is not specified. Their qualifications are stated as "board certified" in their respective fields.
4. Adjudication method (e.g. 2+1, 3+1, none) for the test set
- The document does not describe a specific adjudication method (like 2+1 or 3+1) for establishing the ground truth or evaluating the test set images. It states that experts "reviewed and rated" the images.
5. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance
- No MRMC comparative effectiveness study was done to evaluate human readers' improvement with AI assistance. The studies performed were primarily focused on the standalone performance and safety of the AiLARA algorithm and the updated device.
6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done
- Yes, a standalone evaluation of the AiLARA algorithm was performed. The "AiLARA Algorithm Verification" specifically describes training the model and then sending the full testing set into the model to predict its performance on unseen data, which is a standalone assessment. The "Image Quality Validation Study" and "Radiation Dose Testing" also evaluate the device's output (images and dose) when using the AiLARA modes, which operate without human input on technique factors during image acquisition (the algorithm "determines and recommends a power setting and an exposure time for the X-ray without user input").
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)
- For the "AiLARA Algorithm Verification," the ground truth was referred to as the "truth" which the model was trained to learn the trend of. Given the context of determining optimal technique factors, this "truth" likely relates to ideal exposure parameters for specific anatomies and views.
- For the "Image Quality Validation Study," the ground truth for evaluating image quality was the assessment by board-certified radiologists and an orthopedic surgeon that images were "diagnostically and clinically relevant." This can be considered a form of expert consensus or subjective expert assessment of image utility.
8. The sample size for the training set
- 80% of AiLARA's full development dataset was used for the training set. The total size of the "full development dataset" is not specified.
9. How the ground truth for the training set was established
- The document states that the AiLARA model was trained to "learn the trend of the training dataset (truth)." It implies that the ground truth for the training set consisted of the correct or desired power settings and exposure times for various phantom anatomies and orientations. The specific methodology for establishing this "truth" (e.g., manual expert selection, physical measurements, reference images) for the training data is not detailed but is implicitly linked to generating "a clinically relevant image" while "reduc[ing] overexposures."
§ 892.1720 Mobile x-ray system.
(a)
Identification. A mobile x-ray system is a transportable device system intended to be used to generate and control x-ray for diagnostic procedures. This generic type of device may include signal analysis and display equipment, patient and equipment supports, component parts, and accessories.(b)
Classification. Class II.